Re: [Qemu-devel] xen-unstable qemu-upstream: libxl: error: libxl_qmp.c:440:qmp_next: timeout, libxl_qmp.c:705:libxl__qmp_initialize: Failed to connect to QMP

[Sander Eikelenboom]

Thursday, April 18, 2013, 3:32:37 PM, you wrote:

> Il 18/04/2013 15:27, Stefano Stabellini ha scritto:
>>> > 
>>> >  Any thing that springs to mind ?
>> Nope, sorry.
>> Given the commit that your bisection identified and the error message,
>> it's clear that it's a QMP connection issue from libxl. I would start
>> looking in libxl_qmp.c.
>> Maybe Anthony (CC'ed) has some ideas.

> What host is this?  It is likely a glib or qemu-char bug (hopefully
>  the deadlock that Gerd found on RHEL6) that was latent up to this patch.

The host/Dom0 is Debian squeeze (glibc ) on xen-unstable, using qemu-upstream.
The guest is Debian wheezy.

Previously (older xen-unstable and qemu-upstream (from yesterdays tree)), it 
worked without connecting the console (serial=pty and using "xl create -c") and 
only reverting 0ca5aa4f4c4a8bcc73988dd52a536241d35e5223 from qemu-upstream.

With the latest xen-unstable (764012376a096a45bced88ee2ec1a6c17c6c22c7) and 
qemu-upstream (09dada400328d75daf79e3eca1e48e024fec148d):

Without reverting anything, and using serial=pty and connecting the console at 
boot with "xl create -c", i get:
    xc: info: VIRTUAL MEMORY ARRANGEMENT:
      Loader:        0000000000100000->000000000019c864
      Modules:       0000000000000000->0000000000000000
      TOTAL:         0000000000000000->000000002f800000
      ENTRY ADDRESS: 0000000000100000
    xc: info: PHYSICAL MEMORY ALLOCATION:
      4KB PAGES: 0x0000000000000200
      2MB PAGES: 0x000000000000017b
      1GB PAGES: 0x0000000000000000
    libxl: error: libxl_qmp.c:440:qmp_next: timeout
    libxl: error: libxl_qmp.c:705:libxl__qmp_initialize: Failed to connect to 
QMP
    libxl: error: libxl_qmp.c:440:qmp_next: timeout
    libxl: error: libxl_qmp.c:705:libxl__qmp_initialize: Failed to connect to 
QMP
    libxl: error: libxl_qmp.c:690:libxl__qmp_initialize: Connection error: 
Resource temporarily unavailable
    Daemon running with PID 10174
    xenconsole: Could not read tty from store: No such file or directory
    libxl: error: libxl_exec.c:118:libxl_report_child_exitstatus: console child 
[0] exited with error status 2

and qemu log only shows:
    char device redirected to /dev/pts/17 (label serial0)






With reverting commit 0ca5aa4f4c4a8bcc73988dd52a536241d35e5223, and using 
serial=pty and connecting the console at boot with "xl create -c", i get:

The Guest starts fine.

and qemu log only shows:
    char device redirected to /dev/pts/16 (label serial0)
    main-loop: WARNING: I/O thread spun for 1000 iterations
    xen be: vkbd-0: initialise() failed
    xen be: vkbd-0: initialise() failed
    xen be: vkbd-0: initialise() failed






With reverting commit 0ca5aa4f4c4a8bcc73988dd52a536241d35e5223, and using 
serial=pty and NOT connecting the console at boot with "xl create" (omitting 
the -c), i get:

    Parsing config from /etc/xen/domU/production/security.cfg
    xc: info: VIRTUAL MEMORY ARRANGEMENT:
      Loader:        0000000000100000->000000000019c864
      Modules:       0000000000000000->0000000000000000
      TOTAL:         0000000000000000->000000002f800000
      ENTRY ADDRESS: 0000000000100000
    xc: info: PHYSICAL MEMORY ALLOCATION:
      4KB PAGES: 0x0000000000000200
      2MB PAGES: 0x000000000000017b
      1GB PAGES: 0x0000000000000000
    Daemon running with PID 468

So the guest starts, but it runs very very very very slowly.
When using "xl console", and after i hit a key, i see the kernel boot output 
being printed very slowly with about 1 character per second ...

and qemu log only shows:
    char device redirected to /dev/pts/16 (label serial0)
    main-loop: WARNING: I/O thread spun for 1000 iterations



So it looks like it's a output buffer that's filled to the max that makes it 
grind to a halt.

--
Sander

> Paolo